Using multi-level modelling to understand the determinants of happiness

1. Using multi-level modellingto understand the determinants of happiness Dimitris Ballas Social and Spatial Inequalities Group, Department of Geography, University of Sheffield http://sasi.group.shef.ac.uk/ RES-163-27-1013

2. Outline • Measuring happiness and well-being • Individual-level and contextual factors that may be affecting subjective happiness • Geography of happiness in Britain • Happy People or Happy Places? – a multilevel problem • Concluding comments

3. What is happiness? • Buddhist philosophies • Greece, circa 500 BC • Socrates, Plato  • Aristotle (384-322 BC) • Nichomachean Ethics (350 BC) • http://classics.mit.edu/Aristotle/nicomachaen.html • England, 18th century • Jeremy Bentham (1748 – 1832), the principle of Utility • John Stuart Mill (1806 – 1873) – Utilitarianism • http://www.utilitarianism.com/

4. Can happiness be measured and modelled? A person who has had a life of misfortune, with very little opportunities, and rather little hope, may be more easily reconciled to deprivations than others reared in more fortunate and affluent circumstances. The metric of happiness may, therefore, distort the extent of deprivation in a specific and biased way. (Sen, 1987: 45, my emphasis) Andrew Oswald and colleagues: statistical regression models of happiness measuring the impact of different factors and life events upon human well being World Database of Happiness (Ruut Veenhoven)

5. General Health Questionnaire (1) Have you recently: • Been able to concentrate on whatever you are doing? • Lost much sleep over worry? • Felt that you are playing a useful part in things? • Felt capable of making decisions about things? • Felt constantly under strain? • Felt you could not overcome your difficulties?

6. General Health Questionnaire (2) Have you recently: • Been able to enjoy your normal day-to-day activities? • Been able to face up to your problems? • Been feeling unhappy or depressed? • Been losing confidence in yourself? • Been thinking of yourself as a worthless person? • Been feeling reasonably happy all things considered?

7. Subjective happiness measure: HLGHQ1 This measure converts valid answers to questions wGHQA to wGHQL to a single scale by recoding so that the scale for individual variables runs from 0 to 3 instead of 1 to 4, and then summing, giving a scale running from 0 (the least distressed) to 36 (the most distressed). See Cox, B.D et al, The Health and Lifestyle Survey. (London: Health Promotion Research Trust, 1987).

8. Factors and variables linked to subjective happiness (individual level studies) • Age • Education • Social Class • Income • Marital status/relationships • Employment • Leisure • Religion • Health • Life events and activities

9. Happiness and social comparisons “A house may be large or small; as long as the surrounding houses are equally small it satisfies all social demands for a dwelling. But if a palace arises beside the little house, the little house shrinks to a hovel… [and]… the dweller will feel more and more uncomfortable, dissatisfied and cramped within its four walls.” (Marx, 1847)

10. Region / Metropolitan Area * GHQ: general happiness Crosstabulation % within Region / Metropolitan Area GHQ: general happiness Missing Proxy More than Same as or wild respondent usual usual Less so Much less Total Region / Inner London 4.5% 4.3% 14.4% 66.7% 7.7% 2.4% 100.0% Metropolitan Outer London 2.8% 5.7% 10.6% 68.6% 10.2% 2.1% 100.0% Area R. of South East 2.2% 5.0% 11.9% 70.2% 9.1% 1.6% 100.0% South West 1.7% 3.5% 11.3% 74.1% 8.0% 1.4% 100.0% East Anglia 2.1% 1.3% 10.0% 77.4% 8.5% .8% 100.0% East Midlands 2.2% 1.4% 10.9% 76.0% 8.3% 1.3% 100.0% West Midlands 6.6% 4.6% 11.5% 66.0% 9.9% 1.3% 100.0% Conurbation R. of West Midlands .8% 2.2% 10.7% 73.7% 10.7% 2.0% 100.0% Greater Manchester 1.0% 2.6% 11.1% 75.2% 7.7% 2.4% 100.0% Merseyside .4% 4.7% 9.9% 75.5% 8.6% .9% 100.0% R. of North West 1.3% 4.0% 14.5% 70.7% 8.1% 1.3% 100.0% South Yorkshire 1.0% 1.7% 11.3% 71.0% 13.3% 1.7% 100.0% West Yorkshire 2.7% 2.7% 10.7% 73.9% 8.5% 1.4% 100.0% R. of Yorks & Humberside 1.2% 5.5% 10.1% 76.5% 5.5% 1.2% 100.0% Tyne & Wear .4% 3.8% 14.0% 72.7% 6.8% 2.3% 100.0% R. of North 1.8% 2.3% 10.8% 72.3% 11.5% 1.5% 100.0% Wales 3.9% 1.5% 8.8% 70.9% 12.6% 2.3% 100.0% Scotland 1.8% 2.3% 10.8% 74.0% 9.9% 1.3% 100.0% Total 2.2% 3.4% 11.3% 72.2% 9.2% 1.6% 100.0% Geographies of happiness in Britain Source: The British Household Panel Survey, 1991

11. Research questions : • What are the factors that influence different types of individuals’ happiness? • Is the source of happiness or unhappiness purely personal or do contextual factors matter? (and if they do, to what extent?) • If social comparisons are important, what is the spatial scale at which people make their social comparisons? • Happy People or Happy Places?

12. Research methods: • Regression modelling single level analysis to investigate the association between “subjective happiness” and individual level explanatory variables • Multi-level modelling Assesing variation in happiness at several levels simultaneously

13. Multilevel Analysis World  Nation  Region  DistrictElectoral Wards  Neighbourhood  Household  Individual Multilevel modelling enables the analysis of data with complex patterns of variability – suitable to explore the variability of happiness at different levels

14. Multilevel Analysis World  Nation  Region  DistrictElectoral Wards  Neighbourhood  Household  Individual Multilevel modelling enables the analysis of data with complex patterns of variability – suitable to explore the variability of happiness at different levels

15. Combining Data 1991 & 2001 Census of UK population: 100% coverage fine geographical detail small area data available only in tabular format with limited variables to preserve confidentiality British Household Panel Survey: sample size: more than 5,000 households annual surveys since 1991 individual data more variables than census coarse geography household attrition

16. Modelling happiness and well-being: single level models • Demography • Socio-economic • Health • Social context – interaction variables (e.g. “unemployed or not” dummy variable x “district unemployment rate” variable

19. Modelling happiness and well-being: multilevel (Ballas and Tranmer, 2007) • “Null model” – extent of variation • Socio-economic variables and health – random intercepts • Social context – interaction variables

20. Multi-level modelling (4-levels: region, district, household, individual): “null model”

21. Model 1 variance component estimates

22. All variables – random intercepts

23. All variables and interaction variables – random intercepts

24. Model 2: socio-economic / health characteristics (1)

25. Model 2: socio-economic / health characteristics (2)

26. Model 2: socio-economic / health characteristics (3)

27. Model 3: socio-economic / health and interaction (1)

28. Model 3: socio-economic / health and interaction (2)

29. Model 3: socio-economic / health and interaction (3)

30. Model 3: socio-economic / health and interaction (4)

31. Model 2 and 3 significant main effects (1)

32. Model 2 and 3 significant main effects (2)

36. Random coefficients

37. Conclusions • There are individual variations in happiness • Social context matters • Can explore additional geographical variations using multilevel modelling techniques • Some district level variation in happiness does exist, even after accounting for individual and social contex • Need for longitudinal analysis • Analysis for finer geographical scales (spatial microsimulation and multilevel modelling)